JetBrains MCP codebase graph for AI Assistant

What is JetBrains AI Assistant codebase context?

JetBrains AI Assistant codebase context includes open project files, editor selection, local search, project rules, and tool results available to the assistant. MCP extends that context with external tools, but the team must provide and operate the server that exposes those tools. A Graphify bridge only improves the answer when its graph is fresh and its tool responses stay small enough for the assistant to reason with.

• Project files and editor context provide exact source detail.
• AI Assistant can use MCP tools when configured in the IDE.
• A Graphify graph can answer relationship questions through a bridge.
• Source and tests remain authoritative for implementation behavior.
• MCP adds capability, not automatic trust; returned graph paths still need source verification.

Does Graphify provide a native JetBrains plugin?

No. Graphify CLI 0.9.34 lists installers for Trae, Kiro, Aider, Copilot, Cursor, Codex, Claude, and other agents, but no JetBrains installer. Graphify.homes also has no public MCP endpoint at api.graphify.homes, so the page must not pretend that a one-click hosted connection exists. A separate product with a similar name may advertise a hosted MCP service, but that is not part of this site's runtime.

• Do not run graphify jetbrains install; that command is not supported.
• Do not direct Graphify.homes users to a different hosted Graphify Cloud endpoint.
• Use project artifacts directly when no team-managed MCP bridge exists.
• Treat MCP as team infrastructure, not an implied hosted feature.
• Recheck the installer list after upgrading Graphify before changing this guidance.

How does JetBrains AI Assistant connect to MCP servers?

Install the JetBrains MCP Server plugin, open Settings, and configure AI Assistant's Model Context Protocol settings for the project or IDE. The exact UI and transport support can change by IDE release, so follow the current JetBrains documentation instead of copying a stale screenshot or endpoint. Configure only a server your team controls and verify its tool list before using it on a private repository.

• Use a supported JetBrains IDE release.
• Install the MCP Server plugin from JetBrains Marketplace.
• Configure MCP under AI Assistant settings, not the unrelated server-export page.
• Verify the connection in a project before relying on graph tools.
• Inspect the server's exposed tool names and permissions before enabling it.

What should a Graphify MCP bridge expose?

A useful bridge exposes bounded read-only operations: search nodes, explain a symbol, find a path between nodes, list affected nodes, and return report context. Responses should include file paths, confidence labels, and the graph revision so AI Assistant can verify claims against source. Broad tools such as arbitrary shell execution or unrestricted file reads increase risk without improving the structural query.

• Search for symbols or architectural concepts.
• Explain a node and its immediate neighbors.
• Find call or dependency paths between two nodes.
• Return affected paths and documentation links.
• Never expose write, shell, or arbitrary file-system operations.
• Cap response size and return an explicit stale-graph warning when the revision is old.

When should JetBrains use graph queries instead of local search?

Use graph queries for cross-module dependencies, call paths, subsystem boundaries, and architecture impact. Use JetBrains local search for exact text, current file contents, compiler errors, and recent edits that the graph has not yet indexed. The graph is a navigation layer, while the IDE's compiler integration is the source of truth for type and syntax errors.

• Graph first for relationships among known symbols.
• Search first for exact strings and current diagnostics.
• Open source files before the assistant edits code.
• Use compiler and tests as the final verification layer.
• Re-index or rebuild the graph when source changes invalidate the relationship answer.

How do you secure and refresh a JetBrains graph bridge?

Run the bridge in a trusted local or private environment, authenticate every client, and expose only read-only graph operations. Rebuild the graph after structural changes and include its revision in MCP responses so stale architecture cannot silently guide edits. Store access tokens outside the repository and keep source-derived graph content within the same data boundary as the code.

• Keep graph data on infrastructure approved for the source code.
• Use explicit authentication and least-privilege tool access.
• Log graph queries without storing full sensitive responses.
• Refresh after refactors, interface changes, and dependency upgrades.
• Rotate bridge credentials and remove unused project access when team membership changes.